Multi-scale contour detection model based on fixational eye movement mechanism

Abstract

Physiological evidence has shown that classical receptive field (CRF) responses in the primary visual cortex (V1) can be suppressed by its surrounding region, called the non-classical receptive field (nCRF). Currently, the contour detection model based on the physiological characteristics of the V1 region is mainly used to suppress texture and highlight contour information through the inhibition of nCRF features. However, the effect of eye movement on inhibition is not considered in the inhibition calculation of such models. Inspired by the fixational eye movement (FEyeM) mechanism, we propose a multi-scale contour detection model based on fixational eye movement (MsFem) and the surrounding suppression mechanism. A bank of filters was proposed to simulate the influence of FEyeMs on nCRF, and multi-scale cues were utilized to improve the fine and coarse contour extraction and texture inhibition. The experiments showed that MsFem outperformed some biologically motivated ones in retaining the small-scale target contour information and suppressing the large-scale background textures.